diff --git a/requirements.dev.txt b/requirements.dev.txt index 691c5cd..96dfc0b 100644 --- a/requirements.dev.txt +++ b/requirements.dev.txt @@ -1,2 +1,3 @@ tensorflow-cpu==2.14.1 numpy +Pillow \ No newline at end of file diff --git a/src/arch_arm_mvei.h b/src/arch_arm_mvei.h index 37cc60f..0cc5105 100644 --- a/src/arch_arm_mvei.h +++ b/src/arch_arm_mvei.h @@ -47,14 +47,7 @@ TM_INLINE void tm_dot_prod(mtype_t* sptr, mtype_t* kptr,uint32_t size, sumtype_t } TM_INLINE void tm_dot_prod_pack2(mtype_t* sptr, mtype_t* kptr, uint32_t size, sumtype_t* result) -{ - uint32_t cnt; - sumtype_t sum0 = 0; - sumtype_t sum1 = 0; - mtype_t* kptr0 = kptr; - mtype_t* kptr1 = kptr+size; - cnt = size/PARALLEL_CNT; - +{ uint32_t cnt; int8x16_t vecA; int8x16_t vecB;